How to protect your privacy while using AI tools: Tips from Proton’s CEO
AI tools like ChatGPT, Google Gemini, and Microsoft Copilot are becoming part of daily life for many people. They can help draft emails, summarize documents, or generate ideas. But as these services collect more user data to improve their models, the question of privacy grows more urgent. Is it still possible to use AI without giving away everything about yourself? Proton’s CEO, Andy Yen, says yes — but he also points to one risk that keeps him up at night.
What happened
In a recent interview, Yen spoke about the privacy challenges that come with AI adoption. Proton, the company best known for encrypted email and VPN services, has started offering its own AI-powered tools with a privacy-first design. But Yen’s main worry isn’t about the technology itself — it’s about the lack of clarity in how most AI companies handle user data.
According to Yen, many AI services do not clearly explain what data they collect, how long they store it, or whether your conversations are used to train future models. This lack of transparency makes it difficult for users to make informed decisions.
Why it matters
Every time you type a prompt into a chatbot or upload a document for analysis, that data may be sent to a remote server, processed, and stored. In some cases, companies use that input to fine-tune their models, which means your personal information could be embedded in the system. If you’re sharing sensitive work documents, health questions, or personal stories, that is a real privacy risk.
Regulations like the GDPR and CCPA offer some protection, but they don’t always apply to every tool or every jurisdiction. The burden often falls on the user to understand what’s happening behind the scenes. Yen’s concern is that without clear policies, even careful users can unknowingly give away more than they intend.
What readers can do
You don’t have to stop using AI to protect your privacy. With a few deliberate choices, you can keep your data under your control.
1. Read the privacy policy before using a new AI tool.
It is not exciting, but it is necessary. Look for language about data retention, third-party sharing, and whether your inputs are used for model training. If the policy is vague or absent, that is a red flag.
2. Opt out of training where possible.
Some services allow you to disable the use of your data for model improvement. In ChatGPT, for example, you can turn off chat history and training in the settings. Google Workspace users can adjust their AI data usage controls. Do this before you start using the tool.
3. Use end-to-end encrypted services when you can.
Proton’s own AI features are designed to process data client-side or with end-to-end encryption, so the company cannot read your content. Other privacy-focused alternatives like DuckDuckGo’s AI chat also promise not to store conversations. These are safer options if you are handling sensitive material.
4. Run local AI models for sensitive tasks.
If you are working with confidential data, consider using a local model like Llama or Mistral that runs entirely on your own device. Tools like Ollama make this relatively easy, and you never send anything to the cloud. The trade-off is that local models may be less capable than the latest online versions, so weigh privacy against performance.
5. Avoid sharing personally identifiable information.
Even with strong encryption, it is good practice to remove names, addresses, and other identifiers from your queries. Think of it as a habit, like not writing your password on a sticky note.
Sources
The insights from Proton’s CEO are drawn from a recent interview published by Spiceworks (June 2026). For more detail on Proton’s privacy approach, you can refer to their official blog. Additional guidance on opting out of AI training data collection comes from the privacy policies of OpenAI, Google, and Microsoft, as well as independent privacy audits.